Group differences – within and between – pick a standard please!

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

The debate over at The American Scene on Jim Manzi’s article “Undetermined” is now closed to comments so I couldn’t respond to one of the comments but the beauty of being a blogger is that you can use your own forum to vent your response.

The comment that I desperately wanted to respond to was left by Joe Shipman and reads as follows:

One thing that is established beyond any possibility of scientific doubt, of course, is that the genetic variability in IQ within races is much larger than the variability between races; any ethnic group of nontrivial size will have plenty of smart people and plenty of dumb people, and basing, say, educational policy on group rather than individual characteristics is therefore not only unAmerican but scientifically misguided.

Joe, will you join with me in advocating the complete dismantling of efforts to ameliorate the racial and gender wage gaps that exist, in that they too demonstrate that wage variability is larger within groups than between groups? I hit on this topic a few years ago:

It is important to recognize that most wage inequality occurs within and not between groups. The unweighted average Gini coefficient across all race, gender, and education groups was 0.256 in 1995, over 80 percent of the total Gini. Put another way, if all groups had identical mean wage rates (for example, black male dropouts had the same average wages as white male college graduates) but wages differed within groups as they do today, nearly all the inequality in wage rates would remain.

You know, if it’s unAmerican and unscientific to craft social policy on observed group differences then surely the fact that the variability in Black or Hispanic incomes is greater within their groups than it is between their group and, say, Caucasians or Asians, is an unscientific and unAmerican basis upon which to craft social policy to address the between group differences in income. What’s good for the goose is good for the gander, right Joe?

Labels:

16 Comments

  1. One thing that people who haven’t had a kid in public school don’t realize is how much of educational policy _right now_ is driven by race rather than individual characteristics. It’s not just the explicit affirmative action programs — it’s all the stuff that doesn’t happen, such as “tracking” students by ability into different classes or imposing stricter discipline, because that would lead to “disparate impact” on different races, and thus would make the school district vulnerable to discrimination lawsuits.  
     
    “Disparate impact” legal theory is based on the assumption of racial uniformity and typically sets some expensive barriers to be overcome to get around that assumption.  
     
    I’m all in favor of the schools treating everybody as an individual. It’s the current educational establishment that doesn’t.

  2. it’s all the stuff that doesn’t happen, such as “tracking” students by ability into different classes or imposing stricter discipline, because that would lead to “disparate impact” on different races, and thus would make the school district vulnerable to discrimination lawsuits. 
     
    Do readers also know that the SAT abolished the analogies section as of spring 2005? Analogies are more g-loaded than other verbal items, but they had disparate impact on males vs. females in the undesired direction, so out they went (even though most variability was within females and within males obviously). 
     
    There’s an unfortunate tendency, at least in the US, for 20- and 30-somethings to stay completely out of touch with their eventual replacements. Typically, they don’t revisit this demographic at all, and if so, only 20-odd years down the line when their own kids are in middle or high school. If they only knew what was going on in the meantime, they’d be pretty scared about the future, and thus better motivated to take action to prevent it. 
     
    I tutored for 2 1/2 years before starting graduate school, and even those who do come back into contact with their replacements (like the products of Ed schools) will only indulge in Grumpy Old Man complaints about the music becoming more degenerate, while turning a blind eye to the portentous trends mentioned above. Obligatory mock indignance, I guess you’d call it.

  3. biobehavioral science is too soft. 
     
    sociology is hard enough. 
     
    :-)

  4. Steve says: 
     
    One thing that people who haven’t had a kid in public school don’t realize is how much of educational policy _right now_ is driven by race rather than individual characteristics. 
     
    I think you need a spouse who is a teacher in a public school to actually realize all these things, and it helps if the teacher was largely educated overseas … perhaps in a country where diversity is not an issue :-)

  5. I’m always amazed by how impressed some people are by these within/between population figures. It is so often true of traits that I can’t imagine how anyone could be. There is greater variation in height within each sex than between each sex. Obviously the average difference in height between men and women has no genetic basis.

  6. Razib, how do you imagine we could craft social policy based on observed group differences while remaining fair to individuals who vary substantially from their group’s mean?

  7. JLH:  
     
    Let’s think of two groups of people, each 50% of the population–call them alphas and betas. We want to be fair both to these people as individuals, and to them as groups. Let’s suppose we’re deciding which people to put on the varsity football team.  
     
    Now, suppose that athletic ability, size, strength, speed, skill at football, are distributed in the same way for alphas and betas. That is, learning whether someone is an alpha or beta tells you nothing about how good a football player he is. In this case, we can be fair to individuals and to groups. That is,  
     
    a. We apply the same standards to making the team for alphas and betas. That is, we can say “you’ll be cut if you can’t run 100m in less than 14 seconds” or whatever, and apply it uniformly to everyone. (Real standards would depend on position, of course. Your center doesn’t have to be all *that* fast.)  
     
    b. The proportion of alphas and betas on the varsity team will be statistically about even; there may be a few more alphas one year, and a few more betas another. But there won’t be any kind of persistent or overwhelming advantage for alphas over betas.  
     
    In this world, a coach who discriminated in favor of alphas would end up with that kind of persistently larger-than-expected number of alphas on the team. This is the world that many of our policies today, and nearly all our political rhetoric, are based on. The idea is that groups are more-or-less identical in ability, and so persistent differences in outcome–say, more men in engineering or more Jews in medicine–is based on some kind of discrimination.  
     
    Now, suppose we change just one thing: the groups become unequal. Alphas are one standard deviation to the right of betas in ability. Suddenly, it becomes impossible to be fair both at the individual level and at the group level!  
     
    a. Suppose you apply the same standards to alphas and betas–the same 100m dash time, say. If you do this, you will get different group outcomes. You end up with a lot more alphas on the team than betas. Any objective standards that are the same for members of both groups will lead to this kind of difference in outcome. So individual fairness leads to observed group differences. 
     
    b. Suppose you decide you must end up with the groups having the same number of varsity team members. You can only do this by applying different standards for the two groups–you must require a faster 100m dash time for alphas than for betas. The closest to individual fairness you can get is to make sure that the best alphas make up their half of the football team, and the best betas make up their half.  
     
    So long as groups are different in ability, we can’t get both fairness for groups (in outcomes) and fairness for individuals (in opportunities or requirements for success). But we can still be fair to individuals from both groups, at least on average–in the example above, we require the same 100m dash time from both alphas and betas. Very fast betas will still make the team, based on their performance.  
     
    The social policy problem this raises is that it becomes harder to detect discrimination in areas where we can’t objectively measure group differences. Today there are a lot fewer women than men graduating from engineering departments. Thirty years ago, the same was true of medical schools, but not now. If you are worried about subtle discrimination or discouragement of women, how can you decide if it’s happening in the engineering department? You can be pretty sure it was happening in the medical schools (or the lead-in to those schools) thirty years ago, so you know it’s possible that it’s still happening in the engineering departments today. But you can’t be sure, if you accept that women and men might differ in the distribution of relevant abilities.

  8. BJ and the Bear raises an interesting point: 
     
    There is greater variation in height within each sex than between each sex. Obviously the average difference in height between men and women has no genetic basis. 
     
    On average, males are taller than females yet the male variance is larger than the female variance, so it is not uncommon to find men who are shorter than the majority of all women and men who are taller than all women. 
     
    Hmmm, I wonder what else this pattern applies to?

  9. albatross says: 
     
    Thirty years ago, the same was true of medical schools, but not now. If you are worried about subtle discrimination or discouragement of women, how can you decide if it’s happening in the engineering department? You can be pretty sure it was happening in the medical schools (or the lead-in to those schools) thirty years ago, so you know it’s possible that it’s still happening in the engineering departments today. 
     
    Are you sure you can draw that conclusion? It seems to me that males and females have segregated themselves in medicine and concentrate on different areas. Perhaps there is ongoing discrimination in medical schools, or perhaps preferences are being revealed all around.

  10. Razib, how do you imagine we could craft social policy based on observed group differences while remaining fair to individuals who vary substantially from their group’s mean? 
     
    why are you asking me? it isn’t my post. perhaps i should make the “author” entry larger since people have made this mistake many times…. 
     
    in any case, my position isn’t that between group differences should be used craft social policy per se. but, they should be used as data which needs to be considered when one begins the process of formulating policies which presumably aim to achieve some ends. 
     
    obviously the government isn’t an insurance company; or at least most people don’t believe it should behave that way. some people do believe it should behave that way, but they’re irrelevant since they’re a small enough minority that they’ll never get their way because of hardwired perceptions of what “fairness” is. for an example of what i’m talking about concretely i don’t think many people are realistic about the marginal returns on education for less intelligent people (that is, the group whose IQs are below 100). so the whole “college for everyone” idea is just plain dumb and a waste of money. operationally it probably will only add real value to educators who will get more tax dollars….

  11. Razib, my mistake. I guess since you author the majority of the posts here, I just assumed…. 
     
    But then I have the same question for Tangoman. Over the years I’ve seen posts on GNXP advocating gene therapy, white supremacists advocating segregation, etc. I understand Sailer’s point about public school educational policy being perversely crafted in order the thwart the effects of group differences in IQ. But clearly that’s not a policy that Tangoman would agree with. So I’m just wondering what policies Tangoman would proactively (sorry :) advocate instead of respond to once they are already in place.

  12. So I’m just wondering what policies Tangoman would proactively (sorry :) advocate instead of respond to once they are already in place. 
     
    The point of my post was to critique the comment left by Joe Shipman in which he expresses severe disapproval of the notion of crafting public policy on the basis of group differences, in that he believes it unjust due to the fact that measured variance is greater within the group than between groups. If he holds to this position then I would expect him to also call for efforts to dismantle initiatives which are designed to close the wage gap between men and women or between blacks and whites, in that here too we see that the wage variance is greater within the group of women or blacks than it is between women and men or between blacks and whites. How come most liberals don’t have a problem on this point? 
     
    All I’m asking for is some intellectual consistency. Pick one side and work with it, but don’t condemn one instance of group differences forming the basis of public policy and then embrace another instance formed on the same principles.  
     
    Sure, I understand that when liberals advocate closing wage gaps they feel that they’re on the side of justice and affirming the spirits of the folks that they’re setting out to help by boosting their salaries and these actions justify abandoning principle but in the case of education they can’t achieve the same feel-good dynamics and thus it’s easier to condemn the use of group level factors when formulating public policy. 
     
    That simply doesn’t sit well with me. If the rules of the game allow for group level factors to inform public policy then let’s apply them uniformly and if group factors are deemed unacceptable, then here too we should apply the dictate uniformly. I can live with either approach, but the hypocrisy of embracing group factors in favored instances and then rejecting them in disfavored instances simply rubs me the wrong way.

  13. So I’m just wondering what policies Tangoman would proactively (sorry :) advocate instead of respond to once they are already in place. 
     
    Here’s policy that I would advocate in the field of education: All students, upon entering grade school, and regularly thereafter, are tested on ability to master a lesson plan. The results on such tests determine whether they are placed in a school with a normal school day or one with an extended school day. While this scheme is geared to test the individual student the resulting parsing, based on test results, would most definitely show group differences. 
     
    As the KIPP schools have shown, troubled students can achieve content mastery if they are in school for two hours more per day, go to school on Saturdays, and have their school years extended by an additional month. This added time on task brings these students to the same benchmarks as the students who don’t experience difficulty with the curriculum in the traditional school time frame. The troubled students have slower information uptake rates. They can’t be accommodated in classrooms geared to students with faster information uptake rates and still achieve comparable content mastery. 
     
    Currently we hold to the fantasy that all students can process information at the same, or very similar, rates and can thus achieve mastery of basic principles taught them during the school day. By holding to this fantasy we see disastrous performance disparities across groups. I’d advocate that we abandon the fantasy and focus on educating the children even if it means acknowledging that the students who don’t pass the screening test have to face longer school days, school weeks and school years. At the end the goal is to produce an educated student, not a student with high self-esteem who hasn’t mastered their grade-level content. 
     
    The policy can surely be tweaked to the nth degree to insure that students can transfer between streams as their performance increases or decreases but what will be impossible to hide is the composition of the groups. 
     
    This policy would be a first step in that it probably would have little effect on high achieving students but it would at least address the issue of mastering basic content, which should be achievable for almost every student, but as is clearly the case, this basic goal is not being met.

  14. By law to keep my job here at Stanford University I must spend two hours every two years learning about US anti-discrimination laws. I can’t finish the course (and keep my job) unless I agree it would be illegal for the University Transportation Office to refuse to hire any driver who has been convicted of a felony, because this is presumed to be racial discrimination. 
     
    This inversion of logic makes me lose respect for American laws. I value coed safety far above ex-cons’ rights.

  15. I can’t finish the course (and keep my job) unless I agree it would be illegal for the University Transportation Office to refuse to hire any driver who has been convicted of a felony, because this is presumed to be racial discrimination. 
     
    This doesnt sound right. Can you please cite the regulation, law or court case which has determined that and is current law in 2008?

  16. The state of the art in K-12 education, Direct Instruction, (which, of course, is only practiced in a handful of schools) shows us that thr top 1/3 of students can progress through the curricular sequence at arate 25% faster than the middle group and the bottom 1/3 will progress 25% slower. The amount the bottom group covers can be increased somewhat by providing more instructional time (as TangoMan indicates) and by providing more skilled aides and teachers.

a